Update ConeML profile for 810M release

README.md

@@ -4,7 +4,7 @@ sdk: static
 
 # ConeML
 
-ConeML develops compact domain language models and training methodology focused on capability engineering, corpus design, curriculum design, activation, and evaluation.
+ConeML is an independent model lab developing language models and training methodology focused on capability engineering, corpus design, curriculum design, activation, and evaluation.
 
 The company works upstream of the application layer: first determining whether specialized AI is viable for a measurable workflow, then engineering and evaluating the system around that behavior.
 
@@ -12,7 +12,7 @@ The company works upstream of the application layer: first determining whether s
 
 - Domain-model feasibility and capability analysis
 - Corpus and curriculum engineering
-- Compact model training and adaptation
+- Model training and adaptation
 - Held-out, machine-verified, and calibration-focused evaluation
 - Regression testing, maintenance, and system evolution
 
@@ -20,15 +20,16 @@ The company works upstream of the application layer: first determining whether s
 
 ConeML's public model releases are research artifacts that make measured capability movement and failure boundaries inspectable. The models are evidence; the engineering methodology is the differentiator.
 
-### Current Release: ConeML 348M Gamma
+### Latest Release: ConeML 810M Alpha Family
 
-Gamma is a 348M-parameter, scratch-trained compact language model with a 4,096-token context window, published raw evaluation artifacts, measurable executable-code gains, and deep held-out relational-reasoning results.
+The fourth public ConeML model introduces a scratch-trained 810M family with an 8,192-token context. ConeML 810M Alpha favors executed single-function writing and two-step tasks; ConeML 810M Alpha-Arithmetic records a stronger four-lane arithmetic profile. Both ship as Hugging Face weights plus F16 and tested Q8_0 GGUFs.
 
-- [ConeML 348M Gamma](https://huggingface.co/ConeML/coneml-348m-gamma)
-- [Gamma research notes](https://coneml.com/research/coneml-348m-gamma/)
+- [ConeML 810M Alpha](https://huggingface.co/ConeML/coneml-810m-alpha)
+- [ConeML 810M Alpha-Arithmetic](https://huggingface.co/ConeML/coneml-810m-alpha-arithmetic)
+- [810M release analysis](https://coneml.com/research/coneml-810m-alpha/)
 - [Public release lineage](https://coneml.com/research/)
 
-Public research models are released for non-commercial use under CC BY-NC 4.0 unless stated otherwise.
+The model cards publish task-family-matched strengths beside weak GSM8K and HumanEval results, peer wins, resource assumptions, and known limitations. Public research models are released for non-commercial use under CC BY-NC 4.0 unless stated otherwise.
 
 ## ConeML